Docs / Recipes / LiteLLM

// AGENT FRAMEWORK · LITELLM

LiteLLM

Pattern A · full governance

Pass api_base to litellm.completion (or set the env var) so every call routes through Agentics.

Pattern A — routed through the gateway. Your call travels through Agentics via a base-URL swap, so it is governed (routing, budgets, guardrails) inline and the Proof Ledger writes one verifiable receipt server-side. The receipt hash comes back in the x-agentics-receipt response header.

Install

shell
pip install litellm

Point the framework's OpenAI client at the gateway

Use the openai/ prefix so LiteLLM treats the gateway as an OpenAI-compatible endpoint; OPENAI_API_BASE works too.

Python
import litellm resp = litellm.completion( model="openai/@anthropic/claude-opus-4-8", api_base="https://agentics.you/api/v1", # the only change api_key="ak_…", # AGENTICS_API_KEY messages=[{"role": "user", "content": "hi"}], ) print(resp.choices[0].message.content)

You can also configure it via environment: OPENAI_API_BASE=https://agentics.you/api/v1.

Run one call → see the receipt

Run your normal invocation, then read the receipt for that call off the ledger:

shell
# after your run, the last call's receipt hash: agentics log tail --limit 1 # → shows x-agentics-receipt: <receipt_hash> (verify it at /trust)

Every call the framework makes is now routed, governed, and receipted server-side.

Verify a receipt All recipes →